Texture Perception Using Tactile Sensing Glove Based on PVDF Sensors and Machine Learning

This letter presents a tactile sensing glove based on piezoelectric sensors, embedded electronics, and a machine learning (ML)-based approach for texture discrimination. Various time and frequency features were extracted and evaluated through ML algorithms, including support vector machines (SVM), single-layer feed-forward neural networks, and 1-D convolution neural networks. Six naturalistic surface textures were explored by simply sliding the index finger of seven participants on the surfaces with and without the tactile glove. Results showed that by employing an SVM classifier trained on both time and frequency features, a discrimination accuracy of 87% is achieved while utilizing only two sensors. Furthermore, this discrimination behavior was found to exceed the participants (76.79%) when attempting to discriminate the textures using their somatosensory system. This study demonstrated the capability of the tactile sensing glove in extracting tactile information, opening up interesting perspectives for wearable feedback systems for post-stroke rehabilitation.